fix(流式): 提升SSE稳定性并统一超时配置

- 扩展SSE行长与间隔超时处理，补充keepalive

- 写入失败与超长行时发送错误事件，修复并发释放

- 同步默认配置与示例配置，更新Caddy超时/压缩规则

- 新增OpenAI流式超时与超长行测试

测试: go test ./...

backend/internal/handler/openai_gateway_handler.go

@@ -10,6 +10,7 @@ import (
 	"net/http"
 	"time"
 
+	"github.com/Wei-Shaw/sub2api/internal/config"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/openai"
 	middleware2 "github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
@@ -29,11 +30,16 @@ func NewOpenAIGatewayHandler(
 	gatewayService *service.OpenAIGatewayService,
 	concurrencyService *service.ConcurrencyService,
 	billingCacheService *service.BillingCacheService,
+	cfg *config.Config,
 ) *OpenAIGatewayHandler {
+	pingInterval := time.Duration(0)
+	if cfg != nil {
+		pingInterval = time.Duration(cfg.Concurrency.PingInterval) * time.Second
+	}
 	return &OpenAIGatewayHandler{
 		gatewayService:      gatewayService,
 		billingCacheService: billingCacheService,
-		concurrencyHelper:   NewConcurrencyHelper(concurrencyService, SSEPingFormatNone),
+		concurrencyHelper:   NewConcurrencyHelper(concurrencyService, SSEPingFormatComment, pingInterval),
 	}
 }
 
@@ -124,6 +130,8 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		h.handleConcurrencyError(c, err, "user", streamStarted)
 		return
 	}
+	// 确保请求取消时也会释放槽位，避免长连接被动中断造成泄漏
+	userReleaseFunc = wrapReleaseOnDone(c.Request.Context(), userReleaseFunc)
 	if userReleaseFunc != nil {
 		defer userReleaseFunc()
 	}
@@ -202,6 +210,9 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				log.Printf("Bind sticky session failed: %v", err)
 			}
 		}
+		// 账号槽位/等待计数需要在超时或断开时安全回收
+		accountReleaseFunc = wrapReleaseOnDone(c.Request.Context(), accountReleaseFunc)
+		accountWaitRelease = wrapReleaseOnDone(c.Request.Context(), accountWaitRelease)
 
 		// Forward request
 		result, err := h.gatewayService.Forward(c.Request.Context(), c, account, body)